[DO-NOT-MERGE] Test Victor's patch

Include/internal/pycore_cell.h

@@ -2,6 +2,7 @@
 #define Py_INTERNAL_CELL_H
 
 #include "pycore_critical_section.h"
+#include "pycore_object.h"
 
 #ifdef __cplusplus
 extern "C" {
@@ -42,6 +43,31 @@ PyCell_GetRef(PyCellObject *cell)
     return res;
 }
 
+static inline void
+_PyCell_GetStackRef(PyCellObject *cell, _PyStackRef *value_addr)
+{
+    PyObject *value;
+#ifdef Py_GIL_DISABLED
+    value = _Py_atomic_load_ptr(&cell->ob_ref);
+    if (value != NULL) {
+        if (_Py_IsImmortal(value) || _PyObject_HasDeferredRefcount(value)) {
+            *value_addr = (_PyStackRef){ .bits = (uintptr_t)value | Py_TAG_DEFERRED };
+            return;
+        }
+        if (_Py_TryIncrefCompare(&cell->ob_ref, value)) {
+            *value_addr = _PyStackRef_FromPyObjectSteal(value);
+            return;
+        }
+    }
+#endif
+    value = PyCell_GetRef(cell);
+    if (value == NULL) {
+        *value_addr = PyStackRef_NULL;
+        return;
+    }
+    *value_addr = PyStackRef_FromPyObjectSteal(value);
+}
+
 #ifdef __cplusplus
 }
 #endif

Lib/test/test_free_threading/test_list.py

@@ -3,22 +3,22 @@
 from threading import Thread
 from unittest import TestCase
 
-from test import support
 from test.support import threading_helper
 
 
+NTHREAD = 10
+OBJECT_COUNT = 5_000
+
+
 class C:
     def __init__(self, v):
         self.v = v
 
 
 @threading_helper.requires_working_threading()
 class TestList(TestCase):
-    @support.requires_resource('cpu')
     def test_racing_iter_append(self):
-
         l = []
-        OBJECT_COUNT = 10000
 
         def writer_func():
             for i in range(OBJECT_COUNT):
@@ -34,7 +34,7 @@ def reader_func():
 
         writer = Thread(target=writer_func)
         readers = []
-        for x in range(30):
+        for x in range(NTHREAD):
             reader = Thread(target=reader_func)
             readers.append(reader)
             reader.start()
@@ -44,39 +44,32 @@ def reader_func():
         for reader in readers:
             reader.join()
 
-    @support.requires_resource('cpu')
     def test_racing_iter_extend(self):
-        iters = [
-            lambda x: [x],
-        ]
-        for iter_case in iters:
-            with self.subTest(iter=iter_case):
-                l = []
-                OBJECT_COUNT = 10000
-
-                def writer_func():
-                    for i in range(OBJECT_COUNT):
-                        l.extend(iter_case(C(i + OBJECT_COUNT)))
-
-                def reader_func():
-                    while True:
-                        count = len(l)
-                        for i, x in enumerate(l):
-                            self.assertEqual(x.v, i + OBJECT_COUNT)
-                        if count == OBJECT_COUNT:
-                            break
-
-                writer = Thread(target=writer_func)
-                readers = []
-                for x in range(30):
-                    reader = Thread(target=reader_func)
-                    readers.append(reader)
-                    reader.start()
-
-                writer.start()
-                writer.join()
-                for reader in readers:
-                    reader.join()
+        l = []
+
+        def writer_func():
+            for i in range(OBJECT_COUNT):
+                l.extend([C(i + OBJECT_COUNT)])
+
+        def reader_func():
+            while True:
+                count = len(l)
+                for i, x in enumerate(l):
+                    self.assertEqual(x.v, i + OBJECT_COUNT)
+                if count == OBJECT_COUNT:
+                    break
+
+        writer = Thread(target=writer_func)
+        readers = []
+        for x in range(NTHREAD):
+            reader = Thread(target=reader_func)
+            readers.append(reader)
+            reader.start()
+
+        writer.start()
+        writer.join()
+        for reader in readers:
+            reader.join()
 
 
 if __name__ == "__main__":

Lib/test/test_free_threading/test_monitoring.py

@@ -7,15 +7,14 @@
 import weakref
 
 from sys import monitoring
-from test import support
 from test.support import threading_helper
 from threading import Thread, _PyRLock
 from unittest import TestCase
 
 
 class InstrumentationMultiThreadedMixin:
     thread_count = 10
-    func_count = 200
+    func_count = 50
     fib = 12
 
     def after_threads(self):
@@ -37,14 +36,13 @@ def work(self, n, funcs):
     def start_work(self, n, funcs):
         # With the GIL builds we need to make sure that the hooks have
         # a chance to run as it's possible to run w/o releasing the GIL.
-        time.sleep(1)
+        time.sleep(0.1)
         self.work(n, funcs)
 
     def after_test(self):
         """Runs once after the test is done"""
         pass
 
-    @support.requires_resource('cpu')
     def test_instrumentation(self):
         # Setup a bunch of functions which will need instrumentation...
         funcs = []
@@ -220,29 +218,31 @@ def test_register_callback(self):
         for ref in self.refs:
             self.assertEqual(ref(), None)
 
-    @support.requires_resource('cpu')
     def test_set_local_trace_opcodes(self):
         def trace(frame, event, arg):
             frame.f_trace_opcodes = True
             return trace
 
+        loops = 1_000
+
         sys.settrace(trace)
         try:
             l = _PyRLock()
 
             def f():
-                for i in range(3000):
+                for i in range(loops):
                     with l:
                         pass
 
             t = Thread(target=f)
             t.start()
-            for i in range(3000):
+            for i in range(loops):
                 with l:
                     pass
             t.join()
         finally:
             sys.settrace(None)
 
+
 if __name__ == "__main__":
     unittest.main()

Lib/test/test_free_threading/test_type.py

@@ -5,7 +5,6 @@
 from threading import Thread
 from unittest import TestCase
 
-from test import support
 from test.support import threading_helper
 
 
@@ -97,8 +96,9 @@ def reader_func():
 
         self.run_one(writer_func, reader_func)
 
-    @support.requires_resource('cpu')
     def test___class___modification(self):
+        loops = 200
+
         class Foo:
             pass
 
@@ -108,7 +108,7 @@ class Bar:
         thing = Foo()
         def work():
             foo = thing
-            for _ in range(5000):
+            for _ in range(loops):
                 foo.__class__ = Bar
                 type(foo)
                 foo.__class__ = Foo

Lib/test/test_super.py

@@ -4,7 +4,6 @@
 import threading
 import unittest
 from unittest.mock import patch
-from test import support
 from test.support import import_helper, threading_helper
 
 
@@ -515,10 +514,6 @@ def test___class___modification_multithreaded(self):
         an audit hook.
         """
 
-        if support.Py_GIL_DISABLED:
-            # gh-124402: On a Free Threaded build, the test takes a few minutes
-            support.requires('cpu')
-
         class Foo:
             pass
 
@@ -528,7 +523,7 @@ class Bar:
         thing = Foo()
         def work():
             foo = thing
-            for _ in range(5000):
+            for _ in range(200):
                 foo.__class__ = Bar
                 type(foo)
                 foo.__class__ = Foo

Python/bytecodes.c

@@ -1649,12 +1649,11 @@ dummy_func(
 
         inst(LOAD_DEREF, ( -- value)) {
             PyCellObject *cell = (PyCellObject *)PyStackRef_AsPyObjectBorrow(GETLOCAL(oparg));
-            PyObject *value_o = PyCell_GetRef(cell);
-            if (value_o == NULL) {
+            _PyCell_GetStackRef(cell, &value);
+            if (PyStackRef_IsNull(value)) {
                 _PyEval_FormatExcUnbound(tstate, _PyFrame_GetCode(frame), oparg);
                 ERROR_IF(true, error);
             }
-            value = PyStackRef_FromPyObjectSteal(value_o);
         }
 
         inst(STORE_DEREF, (v --)) {

Tools/tsan/suppressions_free_threading.txt

@@ -41,6 +41,8 @@ race_top:tstate_is_freed
 race_top:type_modified_unlocked
 race_top:write_thread_id
 race_top:PyThreadState_Clear
+# see: https://github.com/python/cpython/issues/117721
+race_top:lock_PyThread_release_lock
 # Only seen on macOS, sample: https://gist.github.com/aisk/dda53f5d494a4556c35dde1fce03259c
 race_top:set_default_allocator_unlocked